Automated Sleep Staging Classification System Based On Convolutional Neural Network Using Polysomnography Signals

Abstract

Manual inspection of sleep staging is one of the critical challenges for the last few years because it is a time-consuming process, laborious task, and also gives a heavy burden to the sleep experts for manipulating the entire sleep recording, which alternatively affects their performance. To overcome this challenging issue, we propose an efficient approach to improve the sleep staging accuracy and multi-class classification for sleep behavior analysis using a One-dimensional Convolutional Neural Network (1D-CNN) model with the inputs of polysomnography signal. Methods: The proposed system considers the base model as 1D-CNN model because of its effectiveness, robust and high accuracy to automatically classify the sleep stages from the different medical-conditioned subjects. The proposed system has used stacking ensemble learning techniques for training and classification to improve the sleep staging accuracy for multi-class classification problems and make the system more robust and accurate predictions. Results: The proposed model has been validated on the SHHS dataset using 10-fold cross-validation techniques. This experimental results for the entire fold have been reported for the 1D-CNN for five to two sleep states classification problems. It has been found that the results of the proposed model are incomparable to the earlier state-of-the-art works in terms of classification accuracy. The proposed model is effective for clinical trials during the diagnosis of different types of sleep related disorders, and it becomes helpful to the sleep experts during the diagnosis process.